Reversible Grayscale Watermarking Using Integer Wavelet Transform and Multi-Objective Genetic Algorithm

Most of the reversible watermarking algorithms currently available select binary image as watermark, grayscale watermark should be converted to binary image before embedding. A novel reversible grayscale watermarking algorithm based on human visual system, integer wavelet transforms and multi-objective genetic algorithm is proposed. Histogram modification is performed on the original image to prevent overflow and underflow. Firstly, Integer wavelet transform is performed on the image, and then the coefficients are partitioned into blocks. In each local block two appropriate pixels are found through pixel gray value. Secondly, according to modulo between the gray value difference of the two pixels and six, the third appropriate pixel is found. Watermark is embedded into image by modifying the gray values of the found pixels. The algorithm directly uses gray level image as watermark, and no preprocessing should be taken before embedding watermark. Human visual system is used to improve invisibility. The algorithm calculates watermarking embedding strength dynamically by normalizing entropy of each block. Genetic algorithm is used to optimize results, which meets the quality and invisibility demand after the watermark embedding. Experimental results demonstrate that the proposed method is feasible and easy to implement, meanwhile it is not only completely reversible, but also effective on the transparency and confidentiality.